Lica

Intro

Lica is presented as a research substrate to train and evaluate AI models as design partners. The site and related blog posts describe a focus on building tools and resources for research labs developing next‑generation design AI. Public materials linked from the GDB blog post include a research paper and a GitHub repository, indicating resources intended for evaluation and development rather than a consumer product offering.

Key Features

• GraphicDesignBench (GDB): a real‑world benchmark composed of 49 tasks that evaluate models across layout, typography, SVG, and animation.
• LICA dataset: a collection described as containing 1.5 million layered graphic design compositions for training and analysis.
• Research artifacts: the GDB post links to a paper and a GitHub repository to support reproducibility and deeper technical inspection.
• Transparency markers: the blog tags include privacy and pricing, though explicit pricing or full privacy-policy details are not present in the available evidence.
• Research findings: benchmark evaluation of frontier models is reported to show that professional graphic design remains largely unsolved, underscoring current limitations.

Who this is for

Lica is oriented to research teams and labs working on design AI who need large, structured datasets and task benchmarks for training and evaluation. Given the research focus and the absence of documented pricing, access, licensing, or full privacy details in the provided evidence, institutions planning to use Lica should prepare to engage with the linked paper and GitHub resources and expect a research‑grade offering rather than a turn‑key commercial product.