UIGEN-T2

Summary

UIGEN-T2 is a 41,200-row dataset of prompt–reasoning–response triples for training models to generate web UIs (HTML + Tailwind CSS) with explicit design reasoning. It was used to train the UIGEN-T2 model, a LoRA fine-tune of Qwen2.5-Coder-7B-Instruct on 50,000 diverse UI examples, covering tasks like Jira-like project management tools, finance dashboards, calendars, chat interfaces, and LinkedIn-style profile pages.

Keywords

ui-generation html tailwind-css code-generation reasoning synthetic web-ui

Topics

Code / UI Generation

Research notes

• Part of the UIGEN series by Tesslate. Reasoning traces generated by a specialized teacher model. The associated UIGEN-T2 model uses PEFT/LoRA rank 128.