Recent developments in text-to-image generation have made it possible to create detailed graphics from natural language descriptions. However, raster graphics used in scientific figures often have low resolutions. To solve this issue, researchers have developed AutomaTikZ, a project that explores the usage of visual languages, such as TikZ, to create scientific figures automatically. The project includes the development of DaTikZ, a large-scale TikZ dataset, as well as the adjustment of the LLaMA language model to produce figures similar to those created by humans. The researchers also introduce CLiMA, an extension of LLaMA that enhances text-image alignment and allows the use of photos as additional inputs. Overall, AutomaTikZ aims to improve the creation of scientific vector graphics using language models and visual languages.
Sort: