New SchemaGUI benchmark evaluates LLM GUI generation capabilities

Researchers have introduced SchemaGUI, a new benchmark designed to evaluate the performance of large language models in generating graphical user interfaces (GUIs). This template-based system synthesizes instructions and references from parameterized schemas, enabling the creation of thousands of annotated tasks without manual labeling. Benchmarking five models, including Qwen3.5 and DeepSeek-R1, the study found that precise spatial control remains a significant challenge, with models struggling with complex layouts despite improvements in schema feasibility. The research also indicated that while larger models improve performance, the generation difficulty is highly sensitive to layout complexity, and employing a 'thinking mode' increases token consumption with only marginal gains in GUI scores. AI

IMPACT This benchmark could drive improvements in LLM-driven GUI generation tools and interfaces.

RANK_REASON The cluster contains a research paper introducing a new benchmark for evaluating LLM capabilities. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• arXiv
• DeepSeek-R1
• Qwen 3.5
• Qwen3 coder 30b
• SchemaGUI

• paper
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →