ProductWebGen: Benchmarking Multimodal Product Webpage Generation
Researchers have introduced ProductWebGen, a new benchmark designed to evaluate the capabilities of multimodal generative models in creating product webpages. The benchmark includes 500 test samples across 13 product categories, each with a source image, visual content instruction, and webpage instruction. Two workflows were compared: an editing-based approach using separate LLMs and image editors, and a unified model (UM)-based approach. The editing-based methods showed stronger performance in webpage instruction following and content appeal, while UM-based methods excelled in visual content instructions. AI
IMPACT Establishes a new evaluation standard for multimodal generative models in e-commerce and marketing content creation.