**Summary: **

This research introduces a novel framework leveraging GPT models to guide non-professional users in creating visually appealing layouts for various design applications. The system allows users to specify canvas size and purpose, such as book covers or brochures, and generates layouts accordingly.

**Insights: **

• Simplifies the graphic design process for non-professionals.
• Employs layout reasoning tasks to train the model on executing layout instructions.
• Uses multimodal input to enhance the design flexibility and accuracy.
• Compared to few-shot GPT-4V models, shows a 12% improvement on the Crello benchmark.

**Importance: **

By automating the layout planning process, this technology can democratize design for non-professionals, potentially transforming the creative industry. It demonstrates AI’s growing capability in understanding complex, multimodal user instructions and translating them into practical outputs.