This study explores the use of variational autoencoders (VAEs) to generate complex polymer topologies with specific properties. The researchers constructed a dataset of 1342 polymers with various architectures, including linear, cyclic, branch, comb, star, and dendrimer structures. They employed a multi-task learning framework that reconstructs and classifies polymer topologies while predicting their dilute-solution radii of gyration. The most effective model, TopoGNN, incorporates both graph and topological descriptor features. This model enables the generation of diverse polymer topologies with target characteristic sizes, which are validated through molecular dynamics simulations. The study also investigates how polymer topology affects rheological properties, such as shear viscosity and viscoelastic moduli, at different concentrations. The results demonstrate the potential of VAEs in generating and characterizing complex polymer topologies, opening new avenues for engineering polymers with tailored properties.This study explores the use of variational autoencoders (VAEs) to generate complex polymer topologies with specific properties. The researchers constructed a dataset of 1342 polymers with various architectures, including linear, cyclic, branch, comb, star, and dendrimer structures. They employed a multi-task learning framework that reconstructs and classifies polymer topologies while predicting their dilute-solution radii of gyration. The most effective model, TopoGNN, incorporates both graph and topological descriptor features. This model enables the generation of diverse polymer topologies with target characteristic sizes, which are validated through molecular dynamics simulations. The study also investigates how polymer topology affects rheological properties, such as shear viscosity and viscoelastic moduli, at different concentrations. The results demonstrate the potential of VAEs in generating and characterizing complex polymer topologies, opening new avenues for engineering polymers with tailored properties.