UI2Code^N: A Visual Language Model for Test-Time Scalable Interactive UI-to-Code Generation
- Repository: https://github.com/zai-org/UI2Code_N
- Paper: https://arxiv.org/abs/2511.08195
UI2Code^N is a visual language foundation model trained through staged pretraining, fine-tuning, and reinforcement learning to achieve foundational improvements in multimodal coding, which unifies three key capabilities: UI-to-code generation, UI editing, and UI polishing. Instead of relying on single-turn paradigms that make little use of iterative visual feedback, UI2Code^N introduces an interactive UI-to-code framework that more accurately reflects real-world workflows and raises the upper bound of achievable performance.
Backbone Model
Our model is built on GLM-4.1V-9B-Base.
Quick Inference
This is a simple example of running single-image inference using the transformers library.
First, install the transformers library:
pip install transformers>=4.57.1
Then, run the following code:
from transformers import AutoProcessor, AutoModelForImageTextToText
import torch
messages = [
{
"role": "user",
"content": [
{
"type": "image",
"url": "https://raw.githubusercontent.com/zheny2751-dotcom/UI2Code-N/main/assets/example.png"
},
{
"type": "text",
"text": "Please generate the corresponding html code for the given UI screenshot."
}
],
}
]
processor = AutoProcessor.from_pretrained("zai-org/UI2Code_N")
model = AutoModelForImageTextToText.from_pretrained(
pretrained_model_name_or_path="zai-org/UI2Code_N",
torch_dtype=torch.bfloat16,
device_map="auto",
)
inputs = processor.apply_chat_template(
messages,
tokenize=True,
add_generation_prompt=True,
return_dict=True,
return_tensors="pt"
).to(model.device)
generated_ids = model.generate(**inputs, max_new_tokens=16384)
output_text = processor.decode(generated_ids[0][inputs["input_ids"].shape[1]:], skip_special_tokens=False)
print(output_text)
See our Github Repo for more detailed usage.
Citation
If you find our model useful in your work, please cite it with:
@article{ui2coden2025,
title = {UI2Code$^{N}$: A Visual Language Model for Test-Time Scalable Interactive UI-to-Code Generation},
author = {Yang, Zhen and Hong, Wenyi and Xu, Mingde and Fan, Xinyue and Wang, Weihan and Gu, Xiaotao and Tang, Jie},
journal = {arXiv preprint arXiv:2511.08195},
year = {2025}
}
- Downloads last month
- 45